How Kameleoon counts conversions
Kameleoon Visitor Data Model
Your campaigns' influence can go beyond the limited framework of a single targeted session.
With our visitor data model, you can follow this influence over several successive sessions. When a variation is displayed to a user, all conversions made by that user are attributed to the variation while the campaign is online.
Let's look at an example. You have launched an A/B experiment on a landing page. This experiment generates a large number of conversions, but these conversions take place during non targeted visits. Variation B reduces the bounce rate from 80% to 40%. However, your sales cycle requires an additional visit to trigger the transaction. This second visit is not targeted because it does not involve the landing page. Yet, your experiment may have had a decisive impact on your visitors, and you would like this to be taken into account in the results provided by Kameleoon. This is possible! With our data model, variation B includes these transactions, and the results page announces it as the winner. So, this model takes into account the possibility that some conversions take longer than others.
Attribution window
However, the influence of a campaign is not unlimited. After a certain period of time, we can reasonably consider that the user's behavior is no longer linked to the variation that was assigned to them: this is the window of influence, or attribution window.
What is it?
The attribution window defines the period of time during which visitors' conversions and transactions are attributed to a given variation. By default, it is set to seven days.
For example, let's say a visitor sees a variation on Monday. From the moment they are targeted and enter the experiment, the attribution window opens. That same visitor then completes a transaction the following Wednesday, but is not targeted again by the experiment. The transaction is counted in the results of the variation. At the end of the seven days, the window closes and this visitor's conversions are considered independent of the variation.
The attribution window's limits depend on your sales cycle and the business model of your site. It can also differ from one campaign to another. Kameleoon lets you define these limits (number of hours/number of days) and adapt them to each situation.
Setting up the attribution window
You can set your attribution window's limits in the Configuration page.
Click Admin > Projects > Configuration (on your site's card).
You will find the attribution window at the bottom of the General tab.
When you launch a campaign (experiment or personalization), the value of the attribution window entered on your site will be used throughout your campaign without the possibility of modification: if you modify the attribution window, the change will not impact your online campaign.
If you set the attribution window to 0 days, 0 hours, Kameleoon only counts conversions for targeted visits.
If the campaign's status changes from online to paused, the window originally assigned remains unchanged, guaranteeing consistent results.
You can find the attribution window associated with one of your campaigns on your Dashboards (in the tooltip gathering all the information of the campaign) and on the results page.
Kameleoon's product recommendations module only counts conversions for targeted visits. We do not have plans to offer extended window attribution for product recommendations, but if it is something that would be of interest, please let us know by contacting your Customer Success Manager.
On the results page
You can find the attribution window associated with your campaigns in the Key information section on the right side of the page.
Filter and breakdown
On the results page of your campaign, the filters and breakdowns have a value per visit. If you apply a breakdown, for example, the results table per goal will list all possible unit values and indicate the number of visitors who have validated this value at least once during the attribution window.
Let's take the case of a visitor who makes three visits on different browsers: the first two on Chrome, the next on Firefox. With a breakdown by browser, this visitor would be present in both the Chrome and the Firefox row of the table.
Changing the date range
Temporal filtering is a bit trickier with attribution windows. When you choose a different time range, the selected range may not include the first day on which your experiment started. If this happens, the selected time frame can include visitors who have been previously targeted by the experiment, even though they were not targeted during the selected time period. So, the number of visitors can be greater than the number of visits.
Custom attribution window
The custom attribution window allows you to define experiment-specific attribution windows, enabling greater flexibility in tracking conversions.
The custom attribution window’s default value initially matches the project’s attribution window. If the project-level attribution window is updated, new experiments will automatically reflect this change. You can customize the attribution window by specifying a duration of up to 45 days and 23 hours. The total duration cannot exceed 45 days.
Using the custom attribution window
- Navigate to the Finalization panel.
- Click General settings > Advanced settings.
- Adjust the days and hours fields using your keyboard or the up/down arrows.
If the attribution window exceeds 45 days, a warning message will appear, and the previous valid configuration is restored.
Usage constraints
You can only modify the custom attribution window when the experiment is in:
- Draft status.
- Planned status with no recorded data (before the first planned launch, for example).
You cannot change the custom attribution window for experiments with the following statuses:
- Online, Paused, Stopped, or Planned with data already recorded.
Stopping or Pausing an experiment also stops or pauses its influence.
For more information on how attribution windows work, refer to this section.
Further reading
If you would like more details on how Kameleoon's statistic engine works, you can read our Statistical paper.